Abstract: A positioning and monitoring system for the objects with long distance, large view field and binocular vision was a three-dimensional coordinate measuring system based on binocular stereo vision principle. There was a common problem that the low correct matching rate of multiple non-feature targets randomly appearing led to the reduced accuracy of target positioning. The key was judging the trigger start time correctly. For the situation of the same target appearing continuously in the same side of the image sequence, or the morphological characteristic of the target changing from large to small or splits, the use of centroid method was to eliminate pseudo targets and detect target centroid position correctly. Then image sequences had time synchronization, and multiple targets realized sparse matching based on polar constraint at the same time. The experiment proves that 9 frame live continuous images contain 8 targets, using the above method to complete the targets detection and automatic positioning, and improve the accuracy rate.